Social network types have been proved to have significant impacts on older population's health out-
comes. However, the existing discoveries are still inconsistent, which may be attributed largely to the
heterogeneous measures and methods scholars used and to the unidirectional causalities presumed in
most research. This study addresses these gaps by using more-refined measures to explore whether the
network types have differential impacts on older adults' health outcomes, and whether a reverse causal
relationship exists between older adults' health conditions and the network types they adopted. Using
data from three recent waves (2005, 2008, and 2012) of the Chinese Longitudinal Healthy Longevity
Survey (n ¼ 4190), we constructed four network types using the K-means clustering method (i.e., diverse,
friend, family, and restricted), and examined their impacts on a variety of health outcomes (i.e., physical,
cognitive, psychological, and overall well-being). Our results demonstrate that there are strong reciprocal
associations between these two factors. On the one hand, a diverse network type yielded the most
beneficial health outcomes as measured by multiple health indicators, and the friend-focused network
type is more beneficial than the family-focused network type in physical outcomes but not in psycho-
logical outcomes. On the other hand, we found that a decrease in all health indicators leads to with-
drawal from more-beneficial network types such as a diversified network type, and a shift to less-
beneficial network types such as family-focused or restricted networks. The understanding of this
reciprocal association could encourage programs designed to enhance healthy aging to focus on
improving the bridging social capital of older adults so that they can break the vicious cycle between
network isolation and poor health conditions.